Lecture 1-3: The Theological Assessment of AI's Philosophical Presuppositions

3.1 Three assumptions, rarely argued for

Three philosophical positions recur, usually without being named or defended, whenever AI discourse moves from describing what a system does to claiming what a system is or could become.

Functionalism is the view that what makes something a mental state — a belief, a desire, an instance of understanding — is not what it is made of, but the functional role it plays: the inputs that produce it, the outputs it produces, and its relations to other mental states. On a strict functionalist view, if a system made of silicon and electricity occupies the same functional role a belief occupies in a human brain — produced by the same kinds of inputs, producing the same kinds of outputs — there is no further fact that could disqualify it from being a belief. The position's appeal, and the reason it is attractive to AI researchers whether or not they use the word, is that it makes mind, in principle, substrate-independent: nothing about silicon as opposed to neurons should matter, if the functional organization is right.

Computationalism narrows this further: cognition, on this view, just is a species of computation — information processing over representations, following rules, in a way that is in principle the same kind of process whether it is carried out by a brain or by a sufficiently sophisticated computer. Where functionalism says substrate does not matter to mentality in general, computationalism makes a more specific claim about what the relevant functional organization is: it is computational, which is exactly the vocabulary already used, unreflectively, to describe what a neural network does.

Physicalism, usually the least examined of the three because it is rarely stated as a distinct claim at all, holds that everything that exists, including whatever a mind is, is ultimately physical — that there is no non-physical soul, spirit, or immaterial substance doing any of the relevant work. Functionalism and computationalism do not strictly require physicalism as a matter of logic, but in practice the three arrive together in most AI discourse, because a discussion that assumes cognition is computation over physical representations has, without argument, already assumed there is nothing else in the picture that a computation could fail to capture.

None of these three positions is absurd, and none is being dismissed here as obviously false. Serious philosophers of mind hold versions of all three, and this course is not equipped, nor is it its purpose, to settle debates in professional philosophy of mind. The point of naming them is narrower: a reader who has not noticed that "the AI understands" already assumes something like functionalism has not actually evaluated a claim — they have absorbed one.

3.2 Searle's Chinese Room: a standing objection

The most influential objection to treating computation, by itself, as sufficient for understanding was posed by philosopher John Searle in a 1980 paper and has generated four decades of continuous response since. Searle asks the reader to imagine a person, who speaks no Chinese, locked in a room with a large rulebook written in English. Chinese characters are passed in through a slot; the person consults the rulebook, which specifies, purely by the shapes of the characters, what characters to pass back out; from outside the room, the exchange looks exactly like a fluent conversation in Chinese. Searle's claim is that the person in the room, manipulating symbols by their shapes according to syntactic rules, understands no Chinese whatsoever, no matter how convincing the conversation looks from outside — and that a computer running a program is doing exactly what the person in the room is doing: manipulating symbols according to syntactic rules, with no understanding of what the symbols mean. If the person in the room does not understand Chinese by following the rulebook, a computer does not understand its output by running its program, regardless of how fluent that output appears.

The argument is aimed precisely at computationalism as defined above: it denies that syntax (rule-governed symbol manipulation) is sufficient for semantics (meaning, understanding, "aboutness"). It is worth stating the strongest functionalist replies fairly, because Searle's argument, however influential, has not gone unanswered. The most prominent reply, the "systems reply," argues that Searle is asking the wrong question by asking whether the person in the room understands Chinese — the person is only one component of a larger system that includes the rulebook and the room, and it is the system as a whole, not the person considered alone, that might be said to understand. A related reply notes that human neurons, considered individually, do not understand anything either, and yet a sufficiently organized system of them does; perhaps a sufficiently organized system of symbol-manipulating rules is no different in kind. These replies are taken seriously in the professional literature and this lecture does not pretend Searle's argument is universally regarded as decisive. What it is is a genuine, unresolved objection that anyone confidently asserting "the AI understands" owes an answer to, and the honest state of the debate, four decades on, is that no answer has commanded consensus.

3.3 Colossians 2:8: testing philosophy, not avoiding it

Colossians 2:8 warns against being "robbed" through "philosophy and vain deceit, after the tradition of men, after the elements of the world, and not after Christ." It would misuse this verse to read it as a blanket suspicion of philosophical reasoning as such — the letter itself reasons carefully and philosophically about the supremacy of Christ in the verses immediately surrounding it, and Christian theology has always done serious philosophical work, often borrowing categories from the very philosophical traditions this verse might seem to warn against. The target named in the verse is specific: philosophy that operates "after the tradition of men... and not after Christ" — that is, a framework of assumptions absorbed uncritically, by cultural inheritance rather than examination, and allowed to set the terms by which everything else, including the claims of Christ, gets evaluated.

That is precisely the danger this lecture has been describing. Functionalism, computationalism, and an unstated physicalism are, in most AI discourse, exactly this kind of "tradition of men" — assumptions absorbed through repeated exposure to a vocabulary, not defended through argument, silently determining in advance what counts as a possible answer to "could a machine have a mind." Colossians 2:8's instruction is not "refuse to think about functionalism." It is "notice that you are being asked to accept a framework, and test it before you let it do the work of deciding your conclusions for you." Applied here: a Christian is free to engage functionalism, computationalism, and physicalism as live philosophical positions, argue with their strongest defenders, and even find some version of one persuasive after real examination. What Colossians 2:8 forecloses is absorbing them by osmosis, through the everyday habit of the word "understands," and only noticing the assumption once its theological consequences have already been drawn.

3.4 Genesis 2:7: an early positive counter-claim

Genesis 2:7 does not enter this argument as a proof-text that settles the dualism-physicalism debate by citation — chapter 3 of this course will show that debate is genuinely contested among serious Christian thinkers, and this lecture is not the place to adjudicate it. What Genesis 2:7 offers here is narrower and, for this lecture's purposes, sufficient: an account, stated as early as the text has anything to say about the origin of a human being, of what a living being is made of and how it came to be one. "Yahweh God formed man from the dust of the ground, and breathed into his nostrils the breath of life; and man became a living soul." Two elements are named, and neither is information: dust, a physical, contingent, ordinary material with no special properties of its own, and the breath of life, given directly by God rather than assembled by any process internal to the dust itself. The text's own account of what made the resulting being "a living soul" is not a description of the dust's internal organization reaching some threshold of complexity. It is a divine act performed on the dust from outside it.

This matters directly against computationalism's central claim, because computationalism's account of a mind is, in the relevant sense, a story about internal organization: sufficiently complex information processing over representations is what a mind is, full stop, wherever it occurs. Genesis 2:7 tells a different kind of story about what made one particular collection of dust into a living being — not organizational complexity crossing a threshold, but a personal, relational act of God. Whatever one concludes about whether a computer could in principle be organized in a way that produces genuine understanding — a question this lecture has deliberately left open, following Searle's unresolved standing objection — Genesis 2:7 insists that the biblical account of what makes something alive, and by extension what makes a human being the kind of thing it is, was never told as a story about information architecture. It was told as a story about dust and breath, given by God to a particular creature. Chapters 2 and 3 of this course will develop this claim considerably further; this lecture's task was only to put it on the table early enough that later theological arguments are not the first place a reader encounters it.

3.5 Why order matters: technical before philosophical

This is the reason the chapter was built to run technical-then-philosophical rather than the reverse. A reader who first meets AI through confident claims that "the AI understands" or "the model knows" has already, before any theological chapter begins, absorbed functionalism and computationalism as background assumptions, without having been offered Searle's objection or Colossians 2:8's warning as a counterweight. Lecture 1-1's insistence that an LLM predicts plausible tokens rather than modeling truth was not merely a technical clarification. It was already, without naming it as such, a challenge to computationalism's easy equation of sophisticated output with understanding. This lecture has now named that challenge explicitly, alongside its strongest opposition, so that chapters 2 through 19 argue against the technology and its philosophical framing as they actually are — genuinely capable, genuinely contested at the level of what their capability amounts to — rather than against a caricature easier to defeat than the real thing.

Summary table